diff --git a/src/brad/planner/beam/query_based.py b/src/brad/planner/beam/query_based.py
index 1a59c243..1790654e 100644
--- a/src/brad/planner/beam/query_based.py
+++ b/src/brad/planner/beam/query_based.py
@@ -137,24 +137,21 @@ async def _run_replan_impl(
         ctx.compute_engine_latency_norm_factor()
 
         beam_size = self._planner_config.beam_size()
-        engines = [Engine.Aurora, Engine.Redshift, Engine.Athena]
         first_query_idx = query_indices[0]
+        first_query = analytical_queries[first_query_idx]
         current_top_k: List[BlueprintCandidate] = []
 
-        # Not a fundamental limitation, but it simplifies the implementation
-        # below if this condition is true.
-        assert beam_size >= len(engines)
-
         # 4. Initialize the top-k set (beam).
-        for routing_engine in engines:
+        for routing_engine in Engine.from_bitmap(
+            planning_router.run_functionality_routing(first_query)
+        ):
             candidate = BlueprintCandidate.based_on(
                 self._current_blueprint, self._comparator
             )
             candidate.add_transactional_tables(ctx)
-            query = analytical_queries[first_query_idx]
             candidate.add_query(
                 first_query_idx,
-                query,
+                first_query,
                 routing_engine,
                 next_workload.get_predicted_analytical_latency(
                     first_query_idx, routing_engine
@@ -187,10 +184,16 @@ async def _run_replan_impl(
             next_top_k: List[BlueprintCandidate] = []
             query = analytical_queries[query_idx]
 
+            # Only a subset of the engines may support this query if it uses
+            # "special functionality".
+            engine_candidates = Engine.from_bitmap(
+                planning_router.run_functionality_routing(query)
+            )
+
             # For each candidate in the current top k, expand it by one
             # query in the workload.
             for curr_candidate in current_top_k:
-                for routing_engine in engines:
+                for routing_engine in engine_candidates:
                     next_candidate = curr_candidate.clone()
                     next_candidate.add_query(
                         query_idx,
diff --git a/src/brad/planner/beam/table_based.py b/src/brad/planner/beam/table_based.py
index d0ba2360..5a2903af 100644
--- a/src/brad/planner/beam/table_based.py
+++ b/src/brad/planner/beam/table_based.py
@@ -102,6 +102,7 @@ async def _run_replan_impl(
             )
 
         # 2. Cluster queries by tables and sort by gains (sum).
+        # TODO: Need to consider functionality when creating clusters.
         clusters = self._preprocess_workload_queries(next_workload)
 
         # Sanity check. We cannot run planning without at least one query in the
diff --git a/src/brad/routing/router.py b/src/brad/routing/router.py
index 7f3b579b..f70e3687 100644
--- a/src/brad/routing/router.py
+++ b/src/brad/routing/router.py
@@ -93,7 +93,7 @@ async def engine_for(
         place_support = self._run_location_routing(query, self._table_placement_bitmap)
 
         # Engine functionality constraints.
-        func_support = self._run_functionality_routing(query)
+        func_support = self.run_functionality_routing(query)
 
         # Get supported engines.
         valid_locations = place_support & func_support
@@ -145,7 +145,7 @@ def engine_for_sync(
         # Ideally we re-implement a sync version.
         return asyncio.run(self.engine_for(query, session))
 
-    def _run_functionality_routing(self, query: QueryRep) -> int:
+    def run_functionality_routing(self, query: QueryRep) -> int:
         """
         Based on the functinalities required by the query (e.g. geospatial),
         compute the set of engines that are able to serve this query.